For the last decade, technology leaders have relied on familiar frameworks to understand where value is created in software. We studied cloud infrastructure, platform ecosystems, network effects, and data moats. Then generative AI arrived and changed the rules.
Suddenly, startups built on top of large language models (LLMs) were reaching millions of users in months. Investors rushed to fund AI applications. Enterprises began deploying AI across customer support, sales, operations, legal workflows, and product development.
Yet one question continued to surface:
Where does value actually accrue in the AI era?
If every company has access to the same foundation models, the same APIs, and increasingly similar capabilities, how can any AI business create a durable competitive advantage?
This is the problem the Supply Chain of Intelligence™ framework was designed to solve.
Created by product leader Anand Arivukkarasu, the framework provides a structured way to analyse AI products, understand defensibility, and identify where long-term value is captured across the modern AI stack. Rather than viewing AI as a collection of disconnected technologies, the framework treats intelligence as a supply chain—one that starts with physical computing resources and ends with business outcomes.
For founders, investors, product managers, and enterprise leaders, the framework offers a practical lens for understanding which AI companies are building sustainable advantages and which are merely riding temporary technology waves.
Why Traditional AI Analysis Falls Short
Many discussions about AI focus almost exclusively on models.
People compare GPT, Claude, Gemini, Llama, and other foundation models. They debate benchmark scores, context windows, and reasoning capabilities.
While these discussions matter, they often miss a larger reality:
Customers do not buy models. They buy outcomes.
A legal team doesn’t purchase a language model. They purchase faster contract reviews.
A sales organisation doesn’t buy embeddings. They buy revenue growth.
A healthcare provider doesn’t buy vector databases. They buy improved patient outcomes.
The Supply Chain of Intelligence™ shifts attention away from the underlying technology and toward the complete journey from computation to business value.
This perspective is especially important because foundational models are increasingly becoming interchangeable. As new models enter the market and competition intensifies, the value created by simply wrapping an API becomes harder to defend.
Understanding the Compute-to-Outcome Concept
At the heart of the framework is a simple but powerful idea:
Every AI product transforms compute into outcomes.
This process can be described as a “Compute-to-Outcome” journey.
Imagine a modern AI application.
A user asks a question.
Behind the scenes, multiple systems activate:
Hardware performs calculations
Models process language
Data is retrieved
Business rules are applied
Workflows are executed
Context is remembered
A result is delivered
Most users never see this complexity.
They only experience the final outcome.
The Supply Chain of Intelligence™ maps every layer involved in this transformation and helps organisations understand where strategic leverage exists.
The Core Idea: Intelligence Behaves Like a Supply Chain
Traditional supply chains move physical goods.
Raw materials become components.
Components become products.
Products reach customers.
The Supply Chain of Intelligence™ argues that AI follows a remarkably similar pattern.
Instead of moving physical materials, it moves intelligence.
The framework suggests that intelligence flows through interconnected layers, with each layer adding value before passing it to the next.
Just as some parts of a manufacturing supply chain capture more profit than others, some layers of the AI stack capture more value than others.
Understanding these layers is essential for building a defensible AI company.
The Major Layers of the Supply Chain of Intelligence™
While the framework contains detailed sub-layers, the higher-level structure can be understood through several core categories.
1. Infrastructure Layer
Everything begins with compute.
This includes:
Silicon
GPUs
Data centers
Cloud infrastructure
Networking systems
Without these resources, intelligence cannot be produced.
Companies such as NVIDIA, AMD, Microsoft, Amazon, and Google operate heavily within this layer.
Infrastructure provides the raw power required for AI systems to function.
However, infrastructure alone rarely owns the final customer relationship.
2. Data Layer
Data serves as the fuel for intelligence.
The framework emphasises that not all data is equal.
Particularly valuable forms include:
Proprietary data
Behavioral data
Outcome data
Domain-specific datasets
Organisations that control unique data assets often possess stronger long-term advantages than those relying solely on public information.
3. Model Layer
This is the layer most people associate with AI.
It includes:
Foundation models
Specialized models
Embedding systems
Reasoning models
These technologies transform raw data into usable intelligence.
However, one of the framework’s most important observations is that models are increasingly becoming commoditised.
As competition expands, model capabilities become more widely available, reducing differentiation.
4. Access Layer
The Access Layer determines how intelligence becomes available to applications.
Examples include:
APIs
Integrations
Tool connections
Enterprise systems
This layer acts as a bridge between intelligence generation and practical business use.
Without effective access mechanisms, even the most advanced models remain inaccessible.
5. Execution Layer
Execution is where intelligence begins performing real work.
Examples include:
Writing content
Analyzing legal documents
Automating customer support
Generating software code
This is often where organisations begin realising measurable business value.
6. Orchestration Layer
As AI systems become more sophisticated, multiple tools must work together.
The orchestration layer manages the following:
Agent coordination
Workflow routing
Context management
Human-in-the-loop processes
This layer becomes increasingly important as organisations deploy autonomous and semi-autonomous AI systems.
7. Surface Layer
The surface layer represents the user experience.
This includes:
Chat interfaces
Copilots
Dashboards
Embedded AI experiences
Many AI companies compete aggressively at this layer because it directly shapes customer perception.
However, user interfaces alone are often easy to replicate.
8. Memory Layer
The Memory Layer is one of the most important concepts in the framework.
Memory captures:
User history
Preferences
Organizational knowledge
Institutional context
Learned behavior
Over time, memory creates switching costs.
An AI system that understands years of customer interactions becomes significantly more valuable than one starting from scratch.
This is why many experts consider memory one of the strongest sources of AI defensibility.
Why AI Defensibility Matters
A major contribution of the Anand Arivukkarasu framework is its focus on AI defensibility.
Historically, software companies built moats through:
Network effects
Brand
Distribution
Proprietary technology
AI changes the equation.
When multiple companies can access the same models, differentiation becomes more difficult.
The framework encourages organisations to ask:
What cannot be copied?
What improves over time?
What becomes more valuable with usage?
What creates switching costs?
Companies that answer these questions successfully are far more likely to sustain competitive advantages.
The Problem of Model Commoditization
One of the most discussed ideas within the framework is model commoditisation.
In simple terms:
Models are becoming utilities.
Just as cloud computing became widely accessible, advanced AI capabilities are increasingly available to everyone.
This trend creates pressure on companies that depend entirely on model access.
If a competitor can swap one model for another with minimal disruption, the underlying model is unlikely to be the source of long-term value.
Instead, defensibility shifts toward the following:
Proprietary data
Workflow integration
Customer memory
Outcome ownership
These layers are harder to replicate and therefore more valuable.
AI Product Management Through the Supply Chain Lens
The framework has significant implications for AI product management.
Traditional product managers often focus on features.
AI product managers must think differently.
They must understand:
Where value originates
How intelligence flows through systems
Which layers are vulnerable
Which layers compound over time
Rather than asking:
“What feature should we build next?”
They ask:
“Which layer of the intelligence supply chain are we strengthening?”
This shift creates more strategic decision-making.
It helps product teams focus on durable advantages rather than temporary capabilities.
Outcome Ownership: The Ultimate Destination
Perhaps the most important lesson from the framework is the concept of outcome ownership.
The highest-value companies are often those that own the final business outcome.
For example:
A model provider sells intelligence.
A workflow platform sells productivity.
An outcome owner sells results.
Customers consistently pay higher premiums for guaranteed outcomes than for underlying technology.
As AI evolves, more value may migrate toward organisations that directly own business decisions, completed tasks, and measurable results.
How Founders Can Use the Framework
Founders can apply the framework in several ways:
Evaluate Defensibility
Determine whether the company possesses assets competitors cannot easily copy.
Identify Weak Layers
Discover where dependency on external platforms creates risk.
Prioritize Investments
Focus resources on layers that create long-term leverage.
Improve Product Strategy
Understand how each product decision affects structural advantage.
How Investors Can Use the Framework
Investors increasingly face difficulty distinguishing sustainable AI businesses from temporary AI wrappers.
The Supply Chain of Intelligence™ provides a framework for evaluating:
Competitive positioning
Moat strength
Revenue durability
Platform risk
Long-term value creation
Rather than asking whether a startup uses AI, investors can ask whether it owns a strategically valuable layer of the intelligence supply chain.
The Growing Importance of Memory and Context
As AI capabilities become widespread, memory may emerge as one of the most powerful forms of competitive advantage.
Every interaction generates context.
Every workflow generates history.
Every organisation accumulates institutional knowledge.
Companies that successfully capture and compound this information create intelligence systems that become more valuable over time.
This is one reason many analysts view memory as a future battleground for AI competition.
Where to Learn More About the Framework
For founders, product leaders, and investors interested in evaluating AI businesses through a strategic lens, the framework continues to gain attention across the AI ecosystem.
Resources such as supplychainofai.com provide deeper insights into the model, including layer analysis, defensibility scoring, and practical applications for modern AI companies. The framework’s emphasis on value accrual, defensibility, memory, and outcome ownership offers a useful alternative to model-centric discussions that dominate today’s AI landscape.
Author : Newell Cramen , Dabid weaver , Gopal Krishnan , Sandra Willman , Sam Israel , Saimon Yosef , David Stewart , Nikkolas Jhon Joseph , Maria Robinson , Juliaim Claren , Alex Christian